Mistral AI, a Paris-based AI startup, has successfully raised €600 million ($640 million) in a new funding round, boosting its valuation to $6 billion. This Series B round was spearheaded by General Catalyst. The funding round follows recent reports indicating the company’s pursuit of $600 million at an estimated $5 billion valuation.
Just six months earlier, Mistral AI completed its Series A funding round, led by Andreessen Horowitz (a16z) and Lightspeed Ventures, which valued the company at $2 billion. The latest round also saw participation from existing investors, including Bpifrance, BNP Paribas, and notable corporate backers such as Nvidia, Samsung Venture Investment Corp., and Salesforce Ventures.
Mistral AI’s appeal stems significantly from its founders’ impressive backgrounds. Timothée Lacroix and Guillaume Lample both emerged from Meta Platforms Inc.’s Paris AI Lab, while Arthur Mensch has experience from DeepMind, IBM Corp.’s AI research division.
Founded in April 2023, Mistral AI has quickly gained recognition for its robust multilingual AI models. The flagship model, Mistral Large, demonstrates proficiency in five languages: French, English, German, Spanish, and Italian. It reportedly trails OpenAI’s GPT-4 in reasoning benchmarks by a mere 10%. The company also offers smaller models, Mistral Small and Mistral Medium, designed for various workloads.
Despite entering a market dominated by giants like Meta Platforms Inc. and OpenAI, Mistral AI has managed to attract significant investor interest. CEO Arthur Mensch noted that the investor enthusiasm is a strong endorsement of the company’s disruptive potential. Mensch remarked to the Financial Times that they had disproven claims that the AI market was impervious to disruption, successfully challenging OpenAI’s business model.
In February, Mistral AI partnered with Microsoft, securing access to Azure’s supercomputing cloud infrastructure for model training and deployment. Concurrently, Microsoft invested €15 million ($16.3 million) in Mistral, with this investment converting to equity in the latest funding round.
Additionally, Mistral AI launched Codestral, a powerful open-source coding language model, at the end of May. Capable of understanding over 80 programming languages, Codestral is a 22 billion-parameter model designed to compete with leading models like Meta’s Llama 3 70B, despite being lighter. It can handle prompts up to 32,000 tokens, surpassing Llama 3 70B’s capacity.
Mistral AI’s rapid growth and substantial funding underscore its potential to innovate and compete in the AI sector, challenging established players and pushing the boundaries of AI capabilities.
